Ranking Size-Adjusted Athleticism of 2016 Tight-End Draft Class

Using a logistic regression model based on historical data of combine results and subsequent pro-bowl selections, it was determined that the probability of achieving a pro-bowl can be estimated from combine height, weight, vertical leap, and forty-yard dash times for any incoming tight-end combine participant.  Here are the numbers for the 2016 draft class:

Name	Hgt	Wgt	Dash	Vert	HighPt	Avg_P	P(PB)
Henry, Hunter	77	255	4.67	31.5	108.5	54.6	3.6%
Sandland, Beau	76	253	4.74	35	111	53.4	3.5%
Hooper, Austin	76	254	4.72	33	109	53.8	2.7%
Braunecker, Ben	75	250	4.73	35.5	110.5	52.9	2.4%
Adams, Jerell	77	247	4.64	32.5	109.5	53.2	2.3%
Vannett, Nick	78	257	4.89	30.5	108.5	52.6	1.2%
Malleck, Ryan	76	247	4.81	34.5	110.5	51.4	1.1%
Williams, Bryce	78	257	4.94	29.5	107.5	52.0	0.7%
Anderson, Stephen	74	230	4.63	38	112	49.7	0.7%
Hemingway, Temarrick	77	244	4.71	30.5	107.5	51.8	0.6%
Morgan, David	76	262	5.02	30	106	52.2	0.5%
McGee, Jake	77	250	4.84	28.5	105.5	51.7	0.4%
Grinnage, David	77	248	4.9	29.5	106.5	50.6	0.3%
Duarte, Thomas	74	231	4.72	33.5	107.5	48.9	0.1%

If any notable tight-end hopefuls are missing that you'd like to evaluate their pro-bowl potential, you can use the calculator that was posted previously.  Since these results are adjusted for size, we can see that the low probabilities are generally a result of this class of tight-ends being slightly undersized in terms of weight.  For example, if Hunter Henry had another 10 pounds of muscle with the same speed and vertical, his numbers would be almost identical to Jason Witten and would therefore possess the greater than 6% predicted probability that tends to associated with pro-bowl selection in our historical data set.

SAVVAGE score: Tight-End Pro-Bowl Odds Calculator

Instructions: Enter the significant combine measures to estimate NFL pro-bowl potential of tight-end, or SAVVAGE score

Height (inches):
Weight (pounds):
40-yard Dash (s):
Vertical Leap (in):



High Point Potential:
Average Momentum:
Estimated Pro-bowl Odds: 1 in
Estimated Pro-bowl Probability: %

More on the data source and modeling methods here

Crossover Athlete Rico Gathers has a pro-bowl ceiling as a tight-end

With a tight-end class that is relatively weak on athleticism, Rico Gathers has become an intriguing prospect.  Still, according to two media sources, Gather's last experience on the grid-iron was in middle school at the tender age of 13.  With so little football background, he's a developmental 7th round pick at best, though some team is surely going to give him a chance as an undrafted free agent.

With no football data on Gathers to draw from, we can only perform some comparisons of raw athleticism to similar recruits with more basketball experience than football.  This is the beauty of using a simple model to project tight-end pro-bowl (PB) status, one that predicted both Tyler Eifert and Travis Kelce as Pro-bowlers.  This model based on size adjusted combine scores suggests that Gathers does possess the raw athleticism indicative of a pro-bowl caliber tight end:

Name	Hgt	Wgt	Dash	Vert	High-Pt	Avg_P	P(PB)
Jimmy Graham	78	260	4.53	38.5	116.5	57.40	55.6%
Rico Gathers	78	273	4.75	34.5	112.5	57.47	32.1%
Jordan Cameron	77	254	4.53	37.5	114.5	56.07	27.1%

All of Gather's numbers are from his recent pro-day, with the exception of his vertical, which was a number taken from a 2015 report.  I pulled combine numbers for Graham and Cameron to serve as a reference as they too played a limited roles on their respective college teams (Miami and USC) after playing multiple years of basketball, an option Gathers might consider if he goes undrafted this week.  If you fail to adjust for weight, Gathers looks bad against the 4.53 and 37+ inch verticals of Graham and Cameron.  However, the additional 13-19 pounds that he carries suggest that averaged more momentum over the course of his 40 yard dash than either Graham or Cameron did.  This suggest that once Gathers gets going, he's going to be harder to stop that either of these basketball playing counterparts.

2016 vs. 2015 Draft

In a feat of pure will, Matt Miller of NFL Draft Scout, has evaluated all 400+ draft-worthy players and assigned them a score between 0 and 10.  He kindly shared his GoogleDoc on Twitter, which allowed me to generate the following breakdown of his grades by year:

Clearly, there's a lot of separation in grades by year, particularly in the middle rounds (2-5).  In fact, player in 2016 are expected to be 0.32 points (99% Confidence: 0.29-0.35) higher than those of a comparable rank in 2015. We could follow a similar procedure for quarterbacks, and we find the mean difference is 0.55 points (99% CI: .34-.76) between the same ranked signal-callers from 2016 and 2015.

As many have observed, this draft does appear to be much deeper than last years.  We'll be working up more positional comparisons in the days leading up to the draft, so follow or check back often.

Potential position change for J.J. Watt

The Houston Texans know that J.J. Watt is one versatile athlete.  Prior to his ill-fated rushing attempt as an H-back in last season's wildcard playoff loss to the Chiefs, Watt had some success catching touchdown passes.  In this moonlighting as a tight-end the 2014-2015 season, Watt caught all three short touchdown passes for a total 4 yards. While Watt was undefended in his first TD against the Raiders, he had defenders draped all over him for both of the other touchdown catches.  How can one argue with this kind of athleticism from the 289 pound defensive back?

While the 2014-2015 season is certainly a small and heavily biased sample of red-zone usage to evaluate Watt's potential as a tight-end, there are more numbers to consider.  An exploratory analysis of size adjusted combine measures in TE success estimated that J.J. Watt would have had a 77% chance of being a pro-bowl tight-end, provided he participated in the combine and entered the league at this position.  This model was built solely from combine events, but does indicate that Watt's ceiling as a tight-end was sky high.

Indeed, Watt's ridiculous 4.84 forty-yard dash at 290 pounds corresponds to a weight-adjusted 40-time even higher than the freakishly-athletic, power-lifting Vernon Davis.  Watt's 37 inch vertical and 77 inch frame correspond to a height-complemented-vertical of 114 inches, which is a half-inch higher than that of pro-bowler Greg Olsen.  This athleticism is jaw-dropping when one considers that Watt weighs 35 lbs more than most pass-catching tight-ends in the NFL.  Also, it's reasonable to speculate that  the force that can propel Watt's nearly 300 lbs over three feet into the air can adequately block an opposing defensive end.  This is particularly true when we consider that Watt is armed with the full knowledge of the arsenal of trick that other DE's in the league are likely to have at their disposal.

While position changes are certainly rare in the NFL, tight-end is hardly foreign to Watt: he played the position as a freshman for the Chippewas of Central Michigan back in 2007.  Granted, Watt only caught a meager 8 passes of 11 targets (72.9% catch rate) for 77 yards and no TDs, but he joked that his limited usage was due to Antonio Brown "hogging all the targets."  Indeed, Brown had over 130 targets in 2007.  Is it any wonder Watt chose to transfer to Wisconsin and walk on as a defensive end?

Clearly this All-Pro is better off signing ridiculous contracts as the dominant Defensive End that he has worked to become for as long as possible.  However, Watt has been quoted saying his retirement will come when he finds competing at DE isn't fun any more, but perhaps he could consider a "semi-retirement" to the other side of the ball?  Maybe the one thing more fun than sacking a quarterback is catching their touchdown passes.  Last I checked, you can see all five of Watt's 2014 touchdowns on NFL.com.  It sure looks like fun to me.

Draft Grades on NFL.com

This post is a bit of a data dump, but I thought it would be nice to take a Bird's Eye View of the Draft Grades (in descending order) according to NFL.com.  The top half of the first round has a steep drop.  Next year we might compare grades to see if there is a measurable difference between the grades in terms of the depth of the draft.

Bases Contributed by 2015 Texas Rangers

This post is exploring an often overlooked statistic in baseball:  base-runner movement following plate appearance in MLB.  Since this data isn't tracked well in the stat sheet, we would have to check the play-by-play.  My source was www.retrosheet.org, which included play-by-play info for all American League Games at the time of data collection with little to no work required (just figuring out how to use the DOS prompt to slice the raw data into the play-by-play appropriately).

When in doubt, root for the home team, so here's the figure for all Texas Rangers with >400 plate appearances in American League Ballparks (due to data limitations of retrosheet).  You might think of "Personal" bases as "from home" as they are all bases to which this player advances himself.  As I didn't want to overthink it, I used plate appearance (PA), rather than at-bats as the denominator so advancement would includes hits, walks and sacrifice situations.

I may work up some additional teams, once I have a more robust statistic developed, as I would like something that factors in both offense (total bases per PA) and defense (total bases allowed per PA).  Still, if you've got a request for a team, post it in the comments section below.

Expected Bases: a new baseball metric

The Runs-Batted-In (RBI) metric is seriously flawed, but on base percentage is not quite perfect either.  If one manages to get on base at the cost of a force-out elsewhere, the value of getting on base is diminished.  For example, let's consider an all-to-common baseball scenario: an otherwise perfect grounder to third results in a force out because of the runners on 1st and 2nd base.  The net effect of such a base hit was no improvement in runner position.  We are in the same situation we started, so this base hit should count for naught.  Even worse, if this grounder led to a double play, we are worse off than we started, even though the batter himself may safely make it to first base.

As a statistician in clinical research, such a situation immediately started me thinking about comparing the change in position of base runners before and after each hit (pre/post measures are my bread and butter).  The trouble with this method of accounting is that if we credit a gained base to the hitter, we must necessarily take it away from someone else if the runner fails to score.  This would naturally fall to the last out of the inning if we stick with our pure pre/post comparisons.  However, this doesn't seem quite fair, so perhaps we ought to split bases lost at the end of the inning equally between all hitters that were responsible for outs.

Breaking down some examples may be instructive.  Let's consider a bases-loaded scenario: a grand-slam advancing first, second and third base-runners home (3+2+1 = 6 bases) and yourself as well (4 bases) would net 10 bases for your team.  Alternatively, ending the inning would lose first, second, and third base (-1-2-3 = -6 bases) by contributing the last out(s).  A triple play to end the inning should bear all the blame for these 6 bases (6 bases lost or -6 bases), but grounding into a double should bear 2/3 of the blame (-4 bases), and a fly, ground, or strike out to end the inning should bear a third of the blame (-2 bases).  Still, not all bases are not created equal, so it's possible this metric can be additionally weighted to reflect the different expected runs from 1st, 2nd, and 3rd respectively.

Such a metric would consider a strike-out, ground-out, or pop-fly that fail to advance base runners as equivalent.  This may seem a bit naive as hitting outcomes seem "better" than striking out.  However, in the long-run, strike-outs will seldom advance base runners (except perhaps with errant pitch or dropped third strike requiring a force out at first), pop-outs will occasionally result in a tag-up advance, and ground-outs may advance base-runners, and so this metric may be able to distinguish between the hitters that advance runners more often than not.  The fact that this single statistic can distinguish between a sacrifice fly that advances a second base runner to 3rd and also a capture a ground-out that advances two runners should be encouraging enough to give it an extra look.